Xiangyu Li 5cc6bddb6e

[Kernel] Add GPTQv2 format support for low-bit or asymmetric quantization, by adapting gptq_gemm (#26092)

2025-10-23 23:26:13 -04:00

..

attention

[Attention] Tune CUTLASS MLA num_splits (#26846)

2025-10-16 06:36:09 -07:00

core

[small][batch invariance] Rename the env and internal flags to simplify usage (#26855)

2025-10-16 21:40:25 +00:00

cpu

…

cutlass_extensions

…

mamba/mamba_ssm

…

moe

Bugfix - pass 'max_num_tokens_padded' into 'moe_lora_align_block_size' (#27311)

2025-10-22 12:23:57 +00:00

quantization

[Kernel] Add GPTQv2 format support for low-bit or asymmetric quantization, by adapting gptq_gemm (#26092)

2025-10-23 23:26:13 -04:00

quickreduce

…

rocm

…

sparse/cutlass

…

activation_kernels.cu

…

cache_kernels.cu

…

cache.h

…

cub_helpers.h

…

cuda_compat.h

…

cuda_utils_kernels.cu

…

cuda_utils.h

…

cuda_view.cu

…

cumem_allocator.cpp

…

custom_all_reduce_test.cu

…

custom_all_reduce.cu

…

custom_all_reduce.cuh

…

custom_quickreduce.cu

…

dispatch_utils.h

…

launch_bounds_utils.h

…

layernorm_kernels.cu

[Chore] Remove unused PolyNorm layer (#27110)

2025-10-17 19:03:43 +00:00

layernorm_quant_kernels.cu

[torch.compile] Enable attention and allreduce fusion without custom ops enabled (#24604)

2025-10-17 08:10:23 -06:00

ops.h

[Kernel] Add GPTQv2 format support for low-bit or asymmetric quantization, by adapting gptq_gemm (#26092)

2025-10-23 23:26:13 -04:00

permute_cols.cu

…

pos_encoding_kernels.cu

…

sampler.cu

[Deepseek v3.2] Remove extra logics in indexer (#26465)

2025-10-21 23:34:03 +00:00

torch_bindings.cpp

[Kernel] Add GPTQv2 format support for low-bit or asymmetric quantization, by adapting gptq_gemm (#26092)

2025-10-23 23:26:13 -04:00

type_convert.cuh

…